Code review comment for lp://qastaging/~ted/libindicator/lp719457

On Sun, 2012-03-04 at 21:34 +0000, charles wrote:
> libindicator doesn't appear to be violating the glib API -- this
> smells like a bug in GHashTable that we'd be better off reporting
> upstream instead of using a local band-aid.
>
> Also, nobody's reported this since March 16 last year, and glib has
> revved several times since then. Maybe it would make sense to CC desrt
> to see if that valgrind log looks familiar?

I actually don't think it's a bug in the hashtable, it's that the remove
in the hashtable us causing an unwatch to be called, which is freeing
the datastructure with the string. It is something that is likely to be
difficult to fix in GDBus as it would have to then queue the destruction
of the data structure until the current call stack returns. Since
that's a non-trivial change, I figured it would be worth working around.